Last year, the film development and production company End Cue produced a short film, called Sunspring, that was entirely written by an artificial intelligence using neural networks. More specifically, it was authored by a recurrent neural network called long short-term memory (LSTM). According to End Cue’s Chief Technical Officer, Deb Ray, the company has come a long way in improving the generative AI aspect of the bot. In this episode, Deb joins host Kyle Polich to discuss how generative AI models are being applied in creative processes, such as screenwriting. Their discussion also explores how data science for analyzing development projects, such as financing and selecting scripts, as well as optimizing the content production process.
As mentioned earlier, “Sunspring” is a 2016 experimental science fiction short film that used an AI to write every word of the film script. The film is about three people living in a futuristic world, possibly on a space station, who are probably in a love triangle. There are three characters: H, who is played by Silicon Valley’s Thomas Middleditch, H2, who is played by Elisabeth Gray, and C who is played by Humphrey Ker. Some critics of the film, which was originally created for Sci-Fi London's 48-Hour Film Competition, describe it as a piece similar to the typical sci-fi B-movie, but has an incoherent plot. However, some film reviews commented on how the film’s incoherence was similar to the enigmatic nonsensicality of Samuel Beckett and David Lynch.
For those who are unfamiliar with End Cue’s work, the company is like a hybrid of a film production company and a software company. One of the unique aspects of End Cue is that the team applies machine learning and deep learning to filmmaking, as well as the film space and production aspect. Deb goes into more details about the film production process -- how film scripts are submitted, evaluated, and, overall, the project management scope of preparing a film script for actual filmmaking. One of the core utilities of an AI is to use it as a tool to try and take away the tedium in the Hollywood industry so that creatives can focus more on their artistic endeavors.
Generating human quality text is a challenging problem because of ambiguity of meaning and difficulty in modeling long-term semantic connections. The bot, Benjamin, that wrote the screenplay for Sunspring is an RNN LSTM that was taught to write screenplays. Benjamin was fed dozens of sci-fi screenplays, including everything from 2001: A Space Odyssey to Cowboys & Aliens to Eternal Sunshine Of The Spotless Mind. However, training using maximum likelihood has its downsides like exposure bias. For example, consider a situation in which during training the prediction of the next word conditioned on the previous word, it becomes infeasible since the previous word may not have been seen in the training data.
For these reasons, End Cue has moved on way beyond the techniques used in Sunspring. According to Deb, the company has improved the generative modeling aspect of the bot. A promising method to improve film screenwriting is using Generative Adversarial Networks (GAN), which is statistical probability model that continuously tries to fool other models, until it can do so with ease. When it finally succeeds, the GAN can generate realistic looking images. In other words, if you show one of these neural networks, say, a million cat images, it can probably do a really good job of recognizing a cat when it sees one. Further, it can generate original cat images all on its own.
So where are we now with AI and screenwriting? And where are we headed?
Deb Ray is slated to speak at the Southern California Data Science Conference on Oct. 22, 2017.